Author |
Message |
Jared Barber (jared_barber)
Member Username: jared_barber
Post Number: 39 Registered: 01-2007
| Posted on Tuesday, January 13, 2009 - 02:32 pm: | |
Hello, I'm not sure what I'm doing wrong. It seems like everything I did was fairly simple so I'm guessing I'm not understanding the whole "region" thing. I have attached the file. You'll see in the "equations" section that I want flexpde to find the pressure drop such that the total flux across my "tube" is 8. I run the code and I report this integral at the end to get that the total flux is 7.5. What's going on? Thanks, Jared |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1209 Registered: 06-2003
| Posted on Tuesday, January 13, 2009 - 05:17 pm: | |
If you vary the Penalty term, you get nicely quantized errors: 7.5 and 7.875. I don't know what causes this. However, if I make this a pseudo-time model with a relaxation equation for Dp: dt(Dp) = 8 - line_integral(u,"middle") it relaxes to the correct solution. I'll try to find out what is going wrong with the steady solution.
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1210 Registered: 06-2003
| Posted on Wednesday, January 14, 2009 - 04:03 pm: | |
There appears to be an error in computing the dependency of a variable on an integral over an imbedded FEATURE. The computed derivative is double the correct value, causing the solution process to merely halve the error on each recomputation. In your problem, this results in a solution sequence for U of 4, 6, 7, 7.5, 7.75, etc. The pseudo-time solution works because it does multiple solution steps. This error will not occur for integrals over external or internal region boundaries, but only in imbedded "Features" with the same region number on both sides. We will correct this error and post a developmental version as soon as we finish testing for other ramifications.
|
|